Changeset 491 in xtideuniversalbios for trunk/XTIDE_Universal_BIOS/Src/Device/IDE

Timestamp:

Dec 15, 2012, 2:46:29 PM (12 years ago)

Author:

krille_n_@…

google:author:

krille_n_@hotmail.com

Message:

Changes:

• Added a new define (USE_UNDOC_INTEL) that enables optimizations possible by using undocumented instructions available on all Intel processors and truly compatible clones. AFAIK the only exceptions are the NEC V-series and the Sony CXQ70108 processors so this option should be safe for use on the AT builds.
• Building BIOSDRVS or the BIOS without MODULE_STRINGS_COMPRESSED would fail due to the recent code exclusions so I changed them a bit. Also fixed the mistaken change to Main.asm
• Changed the Tandy specific info in Configuration_FullMode.txt so it matches the info in the Wiki.
• Optimizations and fixes in general.

Location:

trunk/XTIDE_Universal_BIOS/Src/Device/IDE

Files:

• IdeDmaBlock.asm (modified) (1 diff)
• IdeIO.asm (modified) (4 diffs)
• IdeTransfer.asm (modified) (2 diffs)
• IdeWait.asm (modified) (3 diffs)

trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeDmaBlock.asm

@@ -74 +74 @@
 ;--------------------------------------------------------------------
 TransferBlockToOrFromXTCF:
-    ; 8-bit DMA transfers must be done withing 64k physical page.
+    ; 8-bit DMA transfers must be done within 64k physical page.
     ; XT-CF support maximum of 64 sector (32768 bytes) blocks in DMA mode
     ; so we never need to separate transfer to more than 2 separate DMA operations.

trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeIO.asm

@@ -53 +53 @@
 IdeIO_InputToALfromIdeRegisterInDL:
     xor     dh, dh  ; IDE Register index now in DX
-
+    mov     bx, dx  ; and BX
     mov     al, [di+DPT_ATA.bDevice]
     cmp     al, DEVICE_8BIT_XTIDE_REV2
@@ -63 +63 @@
 
 .InputToALfromMemoryMappedRegisterInDX:
-    mov     bx, JRIDE_COMMAND_BLOCK_REGISTER_WINDOW_OFFSET
-    add     bx, dx
     push    ds
     mov     ds, [di+DPT.wBasePort]  ; Segment for JR-IDE/ISA
-    mov     al, [bx]
+    mov     al, [bx+JRIDE_COMMAND_BLOCK_REGISTER_WINDOW_OFFSET]
     pop     ds
     ret
 
 .ReverseA0andA3fromRegisterIndexInDX:
-    mov     bx, dx
     mov     dl, [cs:bx+g_rgbSwapA0andA3fromIdeRegisterIndex]
     SKIP2B  bx  ; Skip shl dx, 1
 
 .ShlRegisterIndexInDX:
-    shl     dx, 1
+    eSHL_IM dx, 1
     ; Fall to .InputToALfromRegisterInDX
 
@@ -116 +113 @@
 .ShlRegisterIndexInDX:
     add     dl, OFFSET_TO_CONTROL_BLOCK_REGISTERS
-    shl     dx, 1
+    eSHL_IM dx, 1
     jmp     SHORT OutputALtoRegisterInDX
 
@@ -164 +161 @@
 
 .ShlRegisterIndexInDX:
-    shl     dx, 1
+    eSHL_IM dx, 1
     ; Fall to OutputALtoRegisterInDX
 

trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeTransfer.asm

@@ -231 +231 @@
 
     ; Convert ES:SI to physical address
-    xor     dx, dx
-    mov     ax, es
+%ifdef USE_186          ; Bytes EU Cycles(286)
+    mov     ax, es      ; 2     2
+    rol     ax, 4       ; 3     9
+    mov     dx, ax      ; 2     2
+    and     ax, BYTE 0Fh; 3     3
+    xor     dx, ax      ; 2     2
+    add     si, dx      ; 2     2
+    adc     al, ah      ; 2     2
+    mov     es, ax      ; 2     2
+    ;------------------------------------
+                        ; 18    24
+%else ; 808x
+
+%if 0
+                        ; Bytes EU Cycles(808x)
+    mov     al, 4       ; 2     4
+    mov     dx, es      ; 2     2
+    xchg    cx, ax      ; 1     3
+    rol     dx, cl      ; 2     24
+    mov     cx, dx      ; 2     2
+    xchg    cx, ax      ; 1     3
+    and     ax, BYTE 0Fh; 3     4
+    xor     dx, ax      ; 2     3
+    add     si, dx      ; 2     3
+    adc     al, ah      ; 2     3
+    mov     es, ax      ; 2     2
+    ;------------------------------------
+                        ; 21    53
+;
+; Judging by the Execution Unit cycle count the above block of code is
+; apparently slower. However, the shifts and rotates in the block below
+; execute faster than the Bus Interface Unit on an 8088 can fetch them,
+; thus causing the EU to starve. The difference in true execution speed
+; (if any) might not be worth the extra 5 bytes.
+; In other words, we could use a real world test here.
+;
+%endif ; 0
+                        ; Bytes EU Cycles(808x/286)
+    xor     dx, dx      ; 2     3/2
+    mov     ax, es      ; 2     2/2
 %rep 4
-    shl     ax, 1
-    rcl     dx, 1
+    shl     ax, 1       ; 8     8/8
+    rcl     dx, 1       ; 8     8/8
 %endrep
-    add     si, ax
-    adc     dl, dh
-    mov     es, dx
+    add     si, ax      ; 2     3/2
+    adc     dl, dh      ; 2     3/2
+    mov     es, dx      ; 2     2/2
+    ;------------------------------------
+%endif                  ; 26    29/26
     ret
 %endif ; MODULE_8BIT_IDE
@@ -266 +306 @@
 %ifdef MODULE_8BIT_IDE
         dw      IdePioBlock_ReadFrom8bitDataPort    ; 2, DEVICE_8BIT_ATA
-        dw      IdePioBlock_ReadFromXtideRev1       ; 3, DEVICE_8BIT_XTIDE_REV1     
+        dw      IdePioBlock_ReadFromXtideRev1       ; 3, DEVICE_8BIT_XTIDE_REV1
         dw      IdePioBlock_ReadFromXtideRev2       ; 4, DEVICE_8BIT_XTIDE_REV2
         dw      IdePioBlock_ReadFrom8bitDataPort    ; 5, DEVICE_8BIT_XTCF_PIO8

trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeWait.asm

@@ -3 +3 @@
 
 ;
-; XTIDE Universal BIOS and Associated Tools 
+; XTIDE Universal BIOS and Associated Tools
 ; Copyright (C) 2009-2010 by Tomi Tilli, 2011-2012 by XTIDE Universal BIOS Team.
 ;
@@ -10 +10 @@
 ; the Free Software Foundation; either version 2 of the License, or
 ; (at your option) any later version.
-; 
+;
 ; This program is distributed in the hope that it will be useful,
 ; but WITHOUT ANY WARRANTY; without even the implied warranty of
 ; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-; GNU General Public License for more details.      
+; GNU General Public License for more details.
 ; Visit http://www.gnu.org/licenses/old-licenses/gpl-2.0.html
-;       
+;
 
 ; Section containing code
@@ -93 +93 @@
 ;       AL, BX, CX, DX
 ;--------------------------------------------------------------------
-.IdePollBsyAndFlgInAH:
+PollBsyAndFlgInAH:
     call    IdeIO_InputStatusRegisterToAL       ; Discard contents of first read